Release of RAVE
We are happy to announce that we release the code for our RAVE model (see the article) !
It includes RAVE, an autoregressive model to learn RAVE’s prior, and scripts to train them on a custom dataset. Using RAVE you can perform timbre transfer, latent space exploration, conditional and unconditional generation !
We are also currently writing an article to make RAVE go real-time inside max-msp! Coming soon :)
Listen to some unconditional darbouka synthesis made in real-time using max/msp and RAVE ! (thanks Maxime Mantovani for the dataset ! )
About the logo
I was walking around Paris and I came across a sign announcing work on a building, with a very nice RAVE writing ! Could it be a coincidence ?